Tools & software
Beyond the works, Mr. Gattax develops his own infrastructure for AI creation: from a full-stack platform to nodes published for the ComfyUI community.
Platform
An AI creation platform conceived and developed from scratch by Mr. Gattax, from database to interface. More than a product, it is the technical laboratory of the artistic practice: the infrastructure where generation workflows, models, and tools are built, tested, and refined before reaching the works.
Building the infrastructure itself, instead of relying only on ready-made tools, is part of the work: understanding the medium down to the code layer changes what can be done with it.
Open egregoralabs.com →Proprietary upscalers
Three proprietary super-resolution pipelines developed from scratch, each with its own strategy for detail reconstruction, fidelity, and image treatment. Built and refined in real production, they are among the most advanced upscalers available today, accompanied by a set of complementary processing tools inside the platform.
ComfyUI nodes · open source
Five packages published on GitHub in 2025, born from real studio production needs and kept open for the ComfyUI community. Each one solves a concrete pipeline problem: semantic upscaling, regional processing, color, animation, and audio.
A region-based diffusion system for semantic upscaling: it divides the image into areas with their own conditioning, where each region receives a distinct prompt, either manual or generated through VLM captioning. It uses context-aware denoising and smooth blending to maintain coherence, with ControlNet integration for both upscaling and generation from scratch.
View on GitHub →Divides large images into tiles with owner-based padding, processes each part separately, and recombines them with smooth transitions. Deterministic planning and adaptive recombination reduce seams, ghosting, and duplicated detail. Includes algorithm, selection, combination, and mask debug nodes.
View on GitHub →Transfers the palette and color mood of a reference image to a target image while preserving local detail and protecting edges. It combines wavelet analysis, multi-scale chroma transfer, and edge-aware masks to avoid halos and color bleeding, with edge_safety and luma_match controls.
View on GitHub →An animation toolkit based on image lists: batch loader (folders, ZIPs, URLs), multi-folder processor, video frame extractor, AI frame interpolation, and multi-format encoder (GIF, MP4, WebM, AVI, MOV, MKV), with order, FPS, and loop controls.
View on GitHub →Audio enhancement inside ComfyUI: the FlashSR node uses diffusion upsampling for clarity and detail in musical audio, while Fat Llama nodes (GPU/CPU) apply iterative spectral enhancement for sharpness and noise reduction. Includes loudness evaluation, quality metrics, and null-test tools.
View on GitHub →
